AI Playbooks

What playbooks do

While your assistant's persona defines how it speaks, a Playbook defines what it's trying to accomplish. Playbooks are step-by-step instructions that guide your assistant through a specific goal — a repeatable workflow that the AI follows every time it's relevant.

Common uses include:

• Qualifying a sales lead (collecting budget, authority, need, and timeline)
• Troubleshooting a technical issue step by step
• Collecting information for a support ticket
• Booking a meeting or demo

Playbooks make your AI consistent. Without them, the assistant improvises. With them, it follows a proven process every time.

Creating a playbook

Go to AI Assistant > AI Playbooks and click Create Playbook.

Name and description

Give your playbook a clear name that describes its purpose — something like "Lead Qualification" or "Password Reset Troubleshooting." The description is for your team's reference.

Goal

The goal is a single sentence that describes what the assistant is trying to achieve. For example: "Determine if the user is a qualified buyer and collect their name, company, and email address."

The AI uses this goal to stay on track. If the conversation drifts, it will steer back towards the goal.

Instructions

The instructions are the heart of your playbook. They tell the assistant exactly what to do, step by step.

Tips for writing effective instructions:

• Number your steps — Clear sequencing helps the AI follow the flow in order.
• Be specific about what to collect — If you need an email address, say so explicitly. Don't assume the AI will ask for it.
• Define what "done" looks like — Tell the assistant what to do when the goal is met (e.g., "Once you have their email and company size, thank them and let them know a team member will be in touch within one business day.").
• Handle common detours — If a user asks about pricing before giving their email, tell the assistant how to respond (e.g., "Give a brief overview but steer the conversation back to collecting their contact info.").

Versioning

Playbooks are version-controlled, which means you can safely make changes without disrupting live conversations. When you edit a playbook, you create a new Draft version. Once you're happy with the changes, publish it to make it the active version.

You can always view the version history and roll back to a previous version if a change doesn't perform as expected.

Attaching playbooks to assistants

A playbook does nothing on its own — it needs to be attached to an assistant.

1. Go to AI Assistant > Assistants and open the assistant you want to update.
2. Find the Playbooks section in the assistant's configuration.
3. Select the playbook you want to attach.

An assistant can have multiple playbooks attached. The AI determines which one applies based on the customer's intent at the start of the conversation.

A/B testing your playbooks

Not sure which version of a playbook performs better? Use Experiments to run an A/B test. You can compare two versions of a playbook side by side and measure which one results in better outcomes — higher completion rates, fewer handoffs, faster resolution.